Coercive Power Dynamics on Social Media and Threats to Democracy

Allison Koh

Digital Democracy Lab, University of Zurich
December 16, 2022

Current Research

Grey Areas of Digital Transnational Repression on Global Media Platforms

How do globally accessible media attacks targeting diaspora regime critics benefit repressive governments?

They subjugate key actors in transnational information flows.

Against this background, the primary question I address in my dissertation research is, “How do globally accessible media attacks targeting diaspora regime critics benefit repressive governments?”

In the first paper of my dissertation, which outlines a conceptual and empirical basis for studying targeted attacks in this context, I contend that they subjugate key actors in transnational information flows, and that diaspora regime critics pose a unique threat to repressive governments.

The starting point for this paper is that we know from research published widely across academic and non-academic outlets that state repression isn’t just a domestic phenomenon. Repressive governments target prominent individuals in exile and diaspora communities around the world, which include many democracies. We also know from the scholarship on foreign influence operations that governments around the world use influence campaigns to manipulate the domestic politics of a targeted region.

Both of these literatures have overlooked the implications of targeted campaigns that happen “in plain sight” on global platforms, and I argue that a target-centered approach to studying the contours of digital transnational repression on globally accessible platforms is important for understanding the full repertoire of strategies for pro-government actors to curail the public sphere in their favor.

How do globally accessible media attacks targeting diaspora regime critics benefit repressive governments?

They can subtly evade consequences on the global stage.

Platform features
- State-affiliated labels \(\rightarrow\) attribution
Power dynamics and nationalist themes
- Gender: Women* easier to attack
- Language “of the oppressor”
Case study: Toxicity on Pro-CCP Twitter

Another way that targeted media attacks can benefit repressive governments is that they can subtly evade consequences on the global stage. I explore this notion in the second paper of my PhD, which was developed in collaboration with NYU’s Center of Social Media and Politics.

The primary expectation of this paper is that pro-government actors can strategically manipulate platform features and power dynamics with nationalist themes to get away with openly attacking diaspora regime critics. Empirically, we look at state-affiliated media labels on Twitter as a symbol of overt government affiliation. In the absence of these labels, there is some plausible deniability in the attribution of who is producing targeted attacks of diaspora activists.

By power dynamics and nationalist themes, we explore two dimensions. Looking at gender, we argue that male chauvanism and patriarchy are central pillars to upholding authoritarian institutions. Global patriarchal norms also afford pro-government actors more lexical resources for attacking women. Language is also a central pillar of nationalism, and repressive governments (like China) enforce the use their official language in peripheral territories to strengthen regime legitimacy.

We explore these contours of pro-government media attacks in a case study that investigates variations in toxicity (as defined by the Perspective API) on Pro-CCP Twitter, with a focus on attacks targeting the Uyghur diaspora. In a preliminary analysis of a labeled sample of relevant pro-CCP tweets, we find that our expectations on gender uphold (with the average toxicity score of content targeting women to be higher), but that our expectations on language do not. To expand on these preliminary results, we are collecting more information on state-affiliated labels to test our expectations on platform features, and are scaling up analysis by using this initial set of labeled tweets to classify around 85,000 potentially relevant tweets by stance using a BERT-based model and active learning.

How do globally accessible media attacks targeting diaspora regime critics benefit repressive governments?

They can strategically limit the spread of information on domestic state repression.

Adding “noise” to the transnational information environment
In the midst of increased global awareness of domestic repression, pro-CCP attacks…
- \(\uparrow\) coordination
- \(\uparrow\) volume
- \(\downarrow\) topical focus
Setting: Pro-CCP Twitter, ~2013-2022

Finally, in a third empirical paper, I argue that states can use targeted attacks to strategically limit the spread of information on domestic state repression.

When images like the one shown here started spreading globally in 2017, pro-CCP actors engaged in a variety of efforts to control the narrative on repression in the Uyghur region.

While some of the attacks I’ve previously mentioned are rather violent in nature, my overall expectation that targeted attacks are also laden with a lot of noise, and that pro-government actors strategically engage in more “chaos” on social media to manipulate the transnational information environment and confuse the general public on what’s real when it comes to what’s happening in the Uyghur region.

In line with this expectation that noise is strategically employed, I expect that–in the midst of increased global awareness of domestic repression, pro-CCP attacks increase in coordination, increase in overall volume, and decrease in topical focus. These expectations are tested on a corpus of over 2 million potentially relevant tweets that mention key words related to “Xinjiang” (the CCP’s official name for the region) and anything to do with Uyghurs at home and abroad.

Contributions

A target-centered approach to studying pro-government media attacks on the global stage is essential to understanding its implications—especially for democracies.
Diaspora regime critics are uniquely threatening to regimes, which is why they are often targets of pro-government media attacks on global platforms.
This is the first work that applies computational methods to a genre of research that typically uses qualitative methods and interviews.

This dissertation has many theoretical, substantive, and methodological contributions. One of the most important takeaways that I’d like to leave the reader with is that a target-centered approach to studying pro-government media attacks on the global stage is essential to understanding its implications—especially for democracies, where many activists in exile and diaspora communities are located.

Substantively, diaspora regime critics are uniquely threatening to regimes because of their horizontal connections to the domestic opposition and vertical connections to international organizations and media outlets, which is why we’d expect them to be among the most targeted individuals for pro-government actors to attack.

Finally, from a methodological standpoint, this is the first work of its kind that applies computational methods to the study of digital transnational repression, which is a genre of research that typically uses qualitative methods and interviews.

Research Agenda for PRODIGI

Implications of Anti-LGBTQ+ Rhetoric for Digital Democracy

Anti-LGBTQ+ Politics in the Digital Sphere

The global alt-right “blames the gays” to promote their political agendas on social media.

Source: Metro Weekly

:::

As we’ve observed especially in recent weeks in the aftermath of the politically motivated Club Q shooting in Colorado, the global alt-right likes to “blame the gays” to promote their political agendas on social media.

A common thread used to justify Anti-LGBT policies is shown in this meme on the left, in which the alt-right demonizes the LGBT+ community as pedophiles and threats to traditional values in society to justify policies that restrict the freedoms of gender and sexual minorities.

Anti-LBGTQ+ themes are also salient across other policy areas (such as public health and international conflict), as shown in these two images on the right.

SKIP THE REST IF SHORT ON TIME

In the upper right corner here is an Instagram post by Marjorie Taylor Greene, who uses this “groomer” narrative of demonizing the LGBT community to support the notion that they are collectively a public health hazard and are spreading monkeypox to children.

Anti-LGBTQ+ themes also appear in alt-right criticisms of the US military, as shown in this political cartoon on the bottom right. The conservative argument here is that repealing the “Dont Ask Don’t Tell” policy has increased the visibility of the LGBT community in the US military, which weakens its image and structure.

Research Questions

WP1: Theory Development

Which policy frames arise from
online anti-LGBTQ+ rhetoric?

Are some political elites more responsive to
online anti-LGBTQ+ rhetoric than others?

WP2: Computational Methods

Do anti-LGBTQ+ policy frames discussed by U.S. users spread to other countries?

WP3: Policy Responses

Are policy frames with combined online support from disparate groups more likely to spread to the political agenda?

WP4: Effect on Public Opinion

Do social media \(\rightarrow\) political agenda policy frames have a stronger effect on opinions?

Do policy frames with support from disparate groups have a stronger effect on opinions?

The research questions linked to this topic that I am particularly interested in addressing fit very well with the work packages of the PRODIGI project.

Two questions that fit very well with the work package on theory development would be, which policy frames arise from Anti-LGBT rhetoric? and Are political elites more responsive to online anti-LGBT rhetoric than others? I expect that, in addressing this first question that policy frames from this sort of rhetoric span the political spectrum (and are not only salient in alt-right spaces). As for whether some political elites are more responsive to online anti-LGBT rhetoric than others, there’s a more complicated answer to explore here. My initial hunch was that right-wing elites would be more responsive, but of course there’s also a very strong backlash effect that can be observed among other politicians who support the LGBT community.

A research question that would require a lot of creativity on the side of using computational methods would be- Do Anti-LGBT policy frames discussed by US social media users spread to other countries? In line with what’s been observed offline, I’d expect that anti-LGBT policy policy frames discussed by US users do spread to other countries, and it would be a very challenging and interesting research design to figure out how to apply the methods used in the lab’s work on political agenda setting to a transnational, likely multilingual context.

One question that would be pertinent to the work package on policy responses is- Are policy frames with combined online support from disparate groups more likely to spread to the political agenda? The U.S. would be a particularly relevant setting to study this question, although I’d also be interested in exploring this in a more comparative setting.

Finally, a couple of questions that could be interesting to explore in addressing work package four would concern whether anti-LGBTQ+ policy frames that spread from social media to the political agenda have a stronger effect on opinions than other frames do and whether anti-LGBTQ+ policy frames with combined support from disparate groups have a stronger effect on opinions than others do. I think that investigating these sorts of questions would make for an interesting collaboration with the experts on survey experiments in the team, which (if I’m not mistaken) includes Emma.

Collaborations at the University of Zurich

PRODIGI

Theory Development \(\leftrightarrow\) Computational Methods
Policy responses in the U.S.
Content moderation

Zooming out to where I envision potential collaborations at the University of Zurich,

I believe that my main contribution to the PRODIGI project would most closely address work packages 1 and 2, bridging theory development with computational methods (which is in line with what I have done in my dissertation research). Beyond the proposed research area on anti-LGBT+ rhetoric, I’m also very interested in being involved in any ongoing work related to Work Package 3A on policy responses in the U.S., as I have a lot of knowledge on the context and also am interested in leveraging what I learned about repressive governments to study bad actors in democracies and how digital technologies factor into that.

I would also be interested in contributing to the sub-project on Content Moderation; I think that relevant discussions on content moderation among alt-right influencers would be fascinating to look into, especially since at least at a micro-level I’ve observed that some of them are very interested in understanding content moderation policy after their social media activity has been restricted.

Zooming out into the digital democracy lab more generally, my envisioned research in the context of the lab would be most aligned with Fabrizio and co-authors’ work on political agenda setting. I would also love to really explore the possibility of using cross-platform data already maintained by the lab, or perhaps using computational resources to explore YouTube transcripts and alternative platforms that are often used by the alt-right. Telegram would also be a particularly relevant platform for my proposed research.

(for example from press releases and Facebook data

Finally, in this role I’d be interested in being involved in the Digital Society Initiative, espeically the ongoing efforts of the Democracy Community. One other resource I’d be particularly interested in exploring would be the use of social media from the Data Donation Lab. SKIP IF OUT OF TIME I think that using this sort of data could lend itself well to addressing questions related to the targets of the bad actors that I study, in an ethical way that actively involves their consent.

Digital Democracy Lab

Political agenda setting
Cross-platform data

Digital Society Initiative

DSI Community Democracy
Data Donation Lab

Thank you!

koh@hertie-school.org

https://allisonkoh.github.io/

@allisonwkoh@mastodon.social

@allisonkoh_

References

Al-Jizawi, Noura, S. Antis, S. Chan, A. Senft, and R. J. Deibert. 2020. “Digital Transnational Repression.” The Citizen Lab. https://citizenlab. ca/2020/11/annotated ….

Jowett, Garth S., and Victoria O’donnell. 2018. Propaganda & Persuasion. Sage publications.

Ritholtz, Samuel, Allison Koh, and Anita Gohdes. 2022. “Fanning the Flames of Hate: The Transnational Diffusion of Online Anti-LGBT+ Rhetoric and Offline Mobilisation.” GNET.

Coercive Power Dynamics on Social Media and Threats to Democracy

Allison Koh

Current Research

How do globally accessible media attacks targeting diaspora regime critics benefit repressive governments?

How do globally accessible media attacks targeting diaspora regime critics benefit repressive governments?

How do globally accessible media attacks targeting diaspora regime critics benefit repressive governments?

Contributions

Research Agenda for PRODIGI

Anti-LGBTQ+ Politics in the Digital Sphere

Social Media \(\leftrightarrow\) Anti-LGBTQ+ Organizing

Research Questions

Collaborations at the University of Zurich

PRODIGI

Digital Democracy Lab

Digital Society Initiative

Thank you!

References